Product pages describe what is included. This one describes what is not, because in a few cases the absence is the more useful thing to know, and because writing a decision down in public is what stops a later version of us quietly reversing it.
Per person flight risk scoring
A score predicting that an individual employee is likely to leave. It would demonstrate extremely well, and it is asked for by name.
We will not build it. Three reasons, and the third is the one that settles it.
It is acted on quietly, by a manager, in ways the subject never sees. The subject cannot correct it, because they do not know it exists. And it is unfalsifiable: a manager who is told somebody may leave treats them differently, that person leaves, and the score is then recorded as having been right. No accuracy figure captures that, because the measurement changed the thing it was measuring.
What we build instead is the team level version. Where in the organisation are the conditions that historically precede people leaving, reported as a group figure with the contributing factors named and no individual identified anywhere. It answers the same business question, and it cannot be printed and put in somebody's file.
AI screening of job applicants
No résumé scoring. No automated shortlisting. No ranked candidate summaries.
AI in hiring here drafts a job description, and structures a document somebody uploaded so the fields do not have to be typed twice. It does not form an opinion about a person, and it does not put candidates in an order.
This is the largest single feature gap between us and the rest of the market in this area, and we are keeping it.
Automatic actions taken while you are away
An assistant that can act is only ever as safe as the record of what it did and the permission layer that constrained it.
Ours prepares an action and puts it in front of you, pre filled, in the same form with the same validation and the same permission check you would have used yourself. You press submit.
The demonstration that wins is the one where it filled the form in and you sent it. Not the one where it did it overnight and told you afterwards.
Testimonials, customer logos and outcome figures
There are none on this website.
Not because we dislike them, but because we do not yet have the customers that would make them true, and a quotation from a company that is not a customer is a fabricated record. The same applies to a logo wall and to a figure like "a third less admin time", which nobody has measured.
When we have customers who are willing to be quoted, we will quote them by name and they will be real.
Compliance badges
We hold no security certifications, so there are no badges. Each one names an audit that somebody has passed, and displaying one for an audit nobody has performed is the same category of statement as the testimonial above.
When that changes we will name the auditor and the date, so it can be checked.
Deleting data to enforce a limit
Reaching a plan limit does not delete anything. It stops you adding more of the thing you have run out of, and everything already there stays readable, editable and exportable.
A tool that removes your data to enforce its own pricing has confused a commercial boundary with a technical one.
Cross workspace reporting
A report totalling across two workspaces would require a query that reaches across the boundary the entire security model depends on, and then a promise that this one query is only used where it should be.
The supported way to get a combined view is to export from both, which is a deliberate act by somebody who holds access to both, and which leaves a record.
Why publish a list like this
Two reasons.
An absence somebody discovers in month three feels like a missing feature. The same absence stated up front is a position, and it lets you decide now whether it is a position you share.
And the asymmetry runs one way. Shipping one of these and withdrawing it later means telling a paying customer we have removed something, which never happens cleanly. Not shipping it stays reversible.